Independent tephrochronological evidence for rapid and synchronous oceanic and atmospheric temperature rises over the Greenland stadial-interstadial transitions between ca. 32 and 40 ka b2k

(1)

Independent tephrochronological evidence for rapid and synchronous oceanic and atmospheric temperature rises over the Greenland

stadial-interstadial transitions between ca. 32 and 40 ka b2k

Sarah M.P. Berben

^a^,^*

, Trond M. Dokken

^b

, Peter M. Abbott

^c^,^d^,¹

, Eliza Cook

^e

, Henrik Sadatzki

^a^,^b^,^f

, Margit H. Simon

^b

, Eystein Jansen

^a^,^b

aDepartment of Earth Science and Bjerknes Centre for Climate Research, University of Bergen, 5007, Bergen, Norway

bNORCE Norwegian Research Centre and Bjerknes Centre for Climate Research, 5007, Bergen, Norway

cSchool of Earth and Ocean Sciences, Cardiff University, Cardiff, CF10 3AT, United Kingdom

dInstitute of Geological Sciences and Oeschger Center for Climate Research, University of Bern, 3012, Bern, Switzerland

ePhysics of Ice, Climate and Earth, Niels Bohr Institute, University of Copenhagen, 2200, Copenhagen, Denmark

fResearch School of Earth Sciences, Australian National University, Canberra, ACT, 2601, Australia

a r t i c l e i n f o

Article history:

Received 1 November 2019 Received in revised form 10 March 2020 Accepted 11 March 2020 Available online 10 April 2020 Keywords:

Quaternary Paleoclimatology Paleoceanography North Atlantic Sedimentology Marine cores Ice cores

Cryptotephrochronology DO-Events

Synchronization

a b s t r a c t

Understanding the dynamics that drove past abrupt climate changes, such as the Dansgaard-Oeschger (DO) events, depends on combined proxy evidence from disparate archives. To identify leads, lags and synchronicity between different climate system components, independent and robust chronologies are required. Cryptotephrochronology is a key geochronological tool as cryptotephra horizons can act as isochrons linking disparate and/or distant records. Here, we investigated marine sediment core MD99- 2284 from the Norwegian Sea to look for previously identified Greenland ice core cryptotephra horizons and define time-parallel markers between the archives. We explored potential secondary transport and depositional mechanisms that could hamper the isochronous integrity of such horizons. We identified six cryptotephra layers of which four correlate to previously known Greenland ice core horizons.

None of those were identiﬁed in other marine cores and thus, this study contributes greatly to the North Atlantic tephra framework tripling the original amount of existing isochrons between ca. 25 and 60 ka b2k. The latter allow a synchronization between MD99-2284 and the Greenland ice cores between ca. 32 e40 ka b2k, which is, in the North Atlantic, the shortest time-interval during the Last Glacial Period to be constrained by four independent tephra isochrons. Theseﬁndings provide essential tephra-based evidence for synchronous and rapid oceanic and atmospheric temperature rises during the Greenland Stadial-Interstadial transitions. Furthermore, it enables us to estimate the average peak-duration of interstadial temperature overshoots at approximately 136 years. As such, this well-targeted high-resolution investigation successfully demonstrates the use of cryptotephra for geochronological purposes in the marine realm.

1. Introduction

Tephrochronology is considered to be a key correlation tool for

establishing precise synchronization of disparate and/or distant climate archives (e.g.Gr€onvold et al., 1995;Rasmussen et al., 2003;

Wastegård et al., 2006;Lowe et al., 2008;Brendryen et al., 2010;

Lowe, 2011; Blockley et al., 2012; Davies et al., 2012, 2014;

Wastegård and Rasmussen, 2014;Davies, 2015). When geochemically unique volcanic ash (tephra) is ejected during an eruption and rapidly deposited, it can be traced in and between different depositional realms and subsequently, it has the potential to act as an isochron (e.g.Lowe, 2011;Davies, 2015). Isochrons, or time-parallel markers, are used for the independent correlation of climatic sequences which helps to assess the relative timing and phasing of

*Corresponding author.

E-mail addresses:[email protected](S.M.P. Berben),[email protected] (T.M. Dokken),[email protected](P.M. Abbott),[email protected] (E. Cook), [email protected] (H. Sadatzki), [email protected] (M.H. Simon),[email protected](E. Jansen).

1 Now at: Climate and Environmental Physics and Oeschger Centre for Climate Change Research, University of Bern, CH-3012 Bern, Switzerland.

Contents lists available atScienceDirect

Quaternary Science Reviews

j o u r n a l h o me p a g e :w w w .e l se v i e r. co m/ lo ca t e / q u a s c i r e v

https://doi.org/10.1016/j.quascirev.2020.106277

(2)

past events (e.g. Lowe, 2011; Davies, 2015). Determining the phasing of events is crucial for understanding the dynamics con- trolling abrupt climate changes such as the Dansgaard-Oeschger (DO)-events. These events were recorded during the Last Glacial Period in the Greenland ice cores (Bond et al., 1993; Dansgaard et al., 1993) as well as in the North Atlantic marine realm (e.g.

Curry and Oppo, 1997;Dokken and Jansen, 1999) reflecting atmospheric conditions over the Greenland ice-sheet and oceanic vari- ability in the North Atlantic, respectively. However, to obtain a full picture of thefluctuations in, and/or interactions between, the at- mosphere and ocean during these events and therefore, to fully comprehend the dynamics involved, it is crucial to combine and compare the disparate climate archives on independent age models (e.g. Davies et al., 2012). As such, an independent method to correlate these records and test the synchronous or non- synchronous responses (leads/lags) to climatic fluctuations is required and can be provided by tephrochronology.

Within theﬁeld of tephrochronology, the recent expansion in the identiﬁcation of cryptotephra horizons (i.e. low-concentration tephra layers, invisible to the naked eye) offers an increased potential for correlation and integration of different climate archives (Davies, 2015). The Greenland ice cores have been screened for their cryptotephra content over selected intervals (e.g. over rapid climate transitions and around notable volcanic events) between ca. 25 and 45 ka b2k (Davies et al., 2008,2010;Bourne et al., 2015).

Bourne et al. (2015)presents a unique insight into North Atlantic volcanism for that period through a signiﬁcantly enhanced Greenland ice core tephrochronological framework beyond the previously known visible layers (Gr€onvold et al., 1995; Zielinski et al., 1996). The enhancement of the framework is of crucial importance for the synchronization of the Greenland ice cores with other palaeoclimatic records, especially where geochemically unique deposits constrain the transitions between Greenland Sta- dials (GS) and Greenland Interstadials (GI) characterizing DO- events. If the same horizons are traced in distal North Atlantic marine records, these deposits form isochrons that can be used as time-parallel markers linking the different archives.

Correlating tephra horizons depends on robust investigations of the geochemical composition of individual grains within each deposit. Additionally, such information is also useful for assessing depositional processes for marine cryptotephras. In particular, when tephra horizons of multiple eruptions from the same volcano are deposited over a short time-frame, the stratigraphic separation of these events might be hampered. For example, Bourne et al.

(2013) demonstrates that the Faroe Marine Ash Zone III (FMAZ III) deposit, once considered a keystone DO-8 time-parallel marker between the Greenland ice cores (Davies et al., 2010) and a number of marine records (Rasmussen et al., 2003;Wastegård et al., 2006), is a far more complex ash zone than previously thought. The Greenland ice cores record 14 separate Grímsv€otn volcanic eruptions during GS-9 and GI-8 which all fall within the broad compositional range of FMAZ III (Bourne et al., 2013). In order to use one of these ice core tephra horizons as a time-parallel marker, it is necessary to also separate them in the marine realm, which requires a detailed shard proﬁle and geochemical investigation of high sedimentation-rate marine cores (Bourne et al., 2013;Griggs et al., 2014).

For the North Atlantic realm, recent intensive studies focusing on cryptotephra (e.g.Brendryen et al., 2010;Griggs et al., 2014;

Abbott et al., 2016,2018a) resulted in an enhanced North Atlantic marine tephra framework for Marine Isotope Stage (MIS) 2e3 (i.e.

ca. 25 to 60 ka b2k) (Abbott et al., 2018a). Nonetheless, despite this signiﬁcant contribution demonstrating the potential for detecting cryptotephra layers in marine records, the actual number of

isochronous deposits in the North Atlantic (ca. 14) (Abbott et al., 2018a) is profoundly lower than those recorded in the Greenland ice cores (ca. 100) (Bourne et al., 2015). Although several of those North Atlantic deposits possess the potential to act as ice-marine time-parallel markers, only the well-known FMAZ II and NAAZ II horizons have, thus far, been identiﬁed in both marine and ice core records between ca. 25 and 60 ka b2k (Abbott et al., 2018a). The low number of existing isochrons (2) and the discrepancy between the numbers of tephra horizons recorded in the different archives is most likely related to many challenges speciﬁc to the marine realm.

Firstly, detecting low-concentration cryptotephra horizons in marine sediments requires an intensive approach including a sequence of extraction techniques (Davies, 2015). Secondly, various secondary transport and deposition mechanisms can compromise the isochronous nature of marine tephra horizons (e.g.Griggs et al., 2014;Davies, 2015;Abbott et al., 2018b).

The instantaneous deposition of tephra is a fundamental prin- ciple of tephrochronology and thus, it is imperative to fully examine if any processes have delayed transportation and/or reworked shards after initial deposition (e.g.Austin et al., 2004;Brendryen et al., 2010;Griggs et al., 2014;Abbott et al., 2011,2018b). Previ- ous studies demonstrated the added value of assessing the co- variance of ice rafted debris (IRD) and shard concentrations to detect if tephra horizons were transported by ice (Brendryen et al., 2010; Griggs et al., 2014;Abbott et al., 2018b). Employing high- resolution down-core concentration profiles has been shown to be useful for evaluating the influence of sediment reworking through bottom currents and bioturbation (Abbott et al., 2013, 2014;Griggs et al., 2014). A classification scheme for glass shard deposits, recently developed byAbbott et al. (2018b), uses such indicators to classify deposits with common characteristics and explore the influence of primary and secondary processes. These additional steps enable one to disentangle the complex interplay of processes that were active in the North Atlantic during the Last Glacial Period and aid the assessment of the integrity of a deposit as an isochronous marker.Abbott et al. (2018b)additionally assessed the spatial distribution of cryptotephra in the North Atlantic and concluded that areas south and east of Iceland possess the highest potential for preserving isochronous horizons that may expedite the synchronization of palaeoclimatic records.

Here, our overall aim is to stratigraphically resolve individual tephra layers within marine sediment core MD99-2284 that can be correlated with common horizons previously identiﬁed in the Greenland ice cores (Bourne et al., 2013,2015). We conducted a well-targeted high-resolution investigation of marine cryptotephra horizons between DO-5 and DO-9 (i.e. ca. 32e40 ka b2k) and applied the protocol for identiﬁcation, characterization and evaluating depositional controls for marine cryptotephras outlined in Abbott et al. (2018b). MD99-2284, retrieved from the Nordic Seas, was selected due to its high sedimentation-rate, location and the results of prior studies of the sequence. In particular, previous studies presented an age model based on a tuning approach (i.e. an alignment of marine core parameters to ice core parameters) (see 2.1.2.) (Dokken et al., 2013;Sadatzki et al., 2019). Additionally, a wide variety of palaeoclimatic proxy records has already been reconstructed with an exceptionally high temporal resolution (i.e.

decadal to centennial) (see 2.1.2.) (Dokken et al., 2013; Sadatzki et al., 2019). If successful in our initial aim, we intend to 1) syn- chronize marine sediment core MD99-2284 to the Greenland ice cores based on independent cryptotephra time-parallel markers (i.e. isochrons) and thereby, independently verify the existing tuning method and reﬁne the core’s age model, 2) place the marine and ice core proxy records on an integrated chronological framework that constrains rapid climatic events, which further allows for a detailed estimation of the timing and phasing of abrupt events,

(3)

and 3) improve the existing North Atlantic tephra framework through the discovery of previously unidentiﬁed isochrons.

2. Material&methods 2.1. Material

2.1.1. Greenland ice core tephra framework

To construct the most comprehensive MIS 3 tephrochronological framework for Greenland,Bourne et al. (2015)employed an intensive ice sampling methodology for four deep ice cores: North Greenland Ice Core Project (NGRIP) (2917 m asl), North Greenland Eemian Ice Drilling (NEEM) (2484 m asl), Greenland Ice Core Project (GRIP) (3230 m asl) and DYE-3 (2480m asl) (Fig. 1). With the discovery of 73 new tephra deposits, in addition to the previously published 26 horizons (Davies et al., 2010;Bourne et al., 2013), the resulting Greenland ice core tephra framework highlighted the nature and unexpected high frequency of volcanic activity in the North Atlantic region. Several of the tephra horizons were identi- ﬁed in close stratigraphic proximity to sharp transitions that mark abrupt climate events characterizing the Last Glacial Period (i.e.

DO-events). Their stratigraphic positions make these tephra layers key horizons to target and locate within the marine realm (Bourne et al., 2015). For every tephra horizon, the authors assigned an independent age that was derived from the Greenland Ice Core Chronology (GICC05) (Andersen et al., 2006;Svensson et al., 2006).

This Greenland ice core tephra framework strongly increases the opportunities for independent synchronizations between disparate and/or distant climate archives.

Here, the focus lies on those Greenland ice core horizons with strong isochron potential that constrain rapid climatic changes between ca. 32 and 40 ka b2k: 1) the tephra horizons deposited during GI-5, GS-6, GI-6 and GI-7 and 2) the closely spaced eruptive events recorded between GI-8 and GS-10 (Table 1). The geochemical compositions of those layers were compared to the best

available published data to attribute each horizon to the most likely volcanic source (Bourne et al., 2015). The different source volcanoes can be traced back to Iceland and, except for one dacitic/rhyolitic (Hekla) horizon, all tephra layers of interest have a basaltic composition (Table 1). In particular, the layers identiﬁed during both GI-5 and GI-6 originate from Kverkfj€oll (Iceland) but differences in their Al2O3 and TiO2 concentrations allow them to be distinguished one from another. The horizon during GS-6 was attributed to Katla (Iceland) and the closely spaced eruptions between GI-8 and GS-10 are, apart from one, all ascribed to Grímsv€otn (Table 1). Discriminating between the geochemical compositions of these deposits within the same stratigraphic unit in the Greenland ice core record is challenging (Bourne et al., 2013,2015) and even more so in the marine realm (e.g.Griggs et al., 2014;Abbott et al., 2016,2018a). However, due to very small differences in their TiO2

content, it is possible to separate the horizons into different sub- groups despite limited stratigraphic separation (Bourne et al., 2013, 2015).

2.1.2. Marine sediment core MD99-2284

Here, we focus on core MD99-2284 (6222.48 N, 058.81 W) retrieved during the MD114/IMAGES V cruise aboard R/V Marion Dufresne (IPEV) (Dokken et al., 2013) (Fig. 1). This core was taken in the Nordic Seas on the northeasternﬂank of the Faeroe-Shetland channel at a water depth of 1500 m (Dokken et al., 2013).

MD99-2284 is of particular interest as the core site is located along the pathway of inflowing warm Atlantic water into the Nordic Seas via the North Atlantic Current (Fig. 1). Hence, this core was retrieved from an excellent location to capture information with respect to the MIS 3fluctuations of oceanography and sea ice cover. The latter was confirmed by the reconstructed proxy records of near-surface temperature (SST), planktic and benthic foraminiferal isotopes (Dokken et al., 2013) as well as sea ice biomarkers (Sadatzki et al., 2019) that all preserve a record of the GS-GI transitions between DO-5 and DO-9. These marine proxy records

Fig. 1.Map of the northern North Atlantic region presenting the locations of studied core material and Icelandic volcanic systems. Marine sediment core MD99-2284 (6222.48 N, 058.81 W; 1500 m water depth) (green star). Greenland ice cores NEEM, NGRIP, GRIP and DYE-3 (black dots). Icelandic volcanic systems (triangles): Kverkfj€oll (orange), Grímsv€otn (yellow), Katla (blue) and Hekla (black). Surface currents indicate the warm Atlantic water inflow northwards and returning southwards outflow of cold Polar water. NAC¼North Atlantic Current. EGC¼East Greenland Current. Map was made with Ocean Data View (Schlitzer, 2018). (For interpretation of the references to color in thisfigure legend, the reader is referred to the Web version of this article.)

(4)

expose the different conditions during GS and GI intervals as well as the abruptness of the GS-GI transitions in great detail due to the marine core’s exceptionally high sedimentation-rates (ca. 50e400 cm/ka) (Dokken et al., 2013; Sadatzki et al., 2019). The temporal resolution varies between 2.5 and 20 years per cm allowing for a tephra investigation on an extremely high-resolution (i.e. on a decadal time-scale).

Building on a previously published age model for MD99-2284 (Dokken et al., 2013), a more detailed version was presented for the DO-5 to DO-9 time-interval bySadatzki et al. (2019). These age models were constructed by aligning transitions in anhysteretic remanent magnetization (ARM), measured along MD99-2284, with similarﬂuctuations (i.e. DO-signals) in NGRIPd¹⁸O (Andersen et al., 2004,2006;Seierstad et al., 2014). This alignment was based on the work ofKissel et al. (1999)who indicated that the oscillations in magnetic properties in the North Atlantic change in phase with Greenland ice cored¹⁸O records. Moreover, distinct temperature rises in the SST record of MD99-2284, parallel to the ARM increases, were aligned to the stadial-interstadial warming transitions in the NGRIPd¹⁸O record (Sadatzki et al., 2019). Based on change point analysis of the proxy records,Sadatzki et al. (2019)used 11 tie- points and linear interpolation between those for the age model of MD99-2284 (Fig. 2a). This tuning placed the marine sediment core depth-scale on the Greenland Ice Core Chronology (GICC05) (Andersen et al., 2006;Svensson et al., 2006;Seierstad et al., 2014) (Fig. 2a). The tuning approach of Sadatzki et al. (2019)was sup- ported by the positions of the Laschamp and Mono Lake geomag- netic excursions as well as by the identiﬁcation of a tephra horizon (i.e. MD99-2284_3040e3041 cm) that has been linked to the NGRIP 2065.65 m horizon published byBourne et al. (2013). These separate lines of evidence argue for a strong chronological control on the DO-5 to DO-9 time-interval. Thus, the existing chronology of MD99-2284 adds to the value of this marine sediment core to be used for a detailed and high-resolution cryptotephra investigation.

To determinate the depth-intervals where the key tephras from the Greenland ice cores (see 2.1.1.) could be located, the most recent age model ofSadatzki et al. (2019)was used. A wider interval (i.e.

ca. 150 years above and below the selected tephra horizons) was chosen to account for uncertainties in the age-depth model.

Furthermore, the tephra horizon identiﬁed bySadatzki et al. (2019)

was recorded within a low-resolution (every 4 cm) tephra investigation between 3020 and 3180 cm. As this depth-interval stratigraphically corresponds to the separated horizons in the ice cores associated with the marine FMAZ III deposit (Bourne et al., 2013) (see 2.1.1.), a more detailed investigation (i.e. high-resolution quantiﬁcation of shards) of this stratigraphic unit was executed in order to assess whether the stratigraphically separated horizons in the ice cores can be resolved in the marine realm.

Subsequently, a cryptotephra investigation was executed for the following four selected depth-intervals: 1) 2108e2148 cm, 2) 2270e2350 cm, 3) 2645e2700 cm, and 4) 3020e3180 cm, representing the GS-6/GI-5, GS-7/GI-6, GS-8/GI-7 and GS-9/GI-8 transitions, respectively (see numbered boxes inFig. 2). Depending on the composition of the Greenland ice core tephra horizons that were targeted, the tephra shard quantiﬁcation was focused on basaltic shards for depth-intervals one, two and four and on rhyolitic shards for depth-interval three (Fig. 2).

2.2. Methods

The methodology for the cryptotephra investigation in this study follows the protocol outlined byAbbott et al. (2018b)building upon previous studies of marine and terrestrial sequences (e.g.

Pilcher and Hall, 1992;Turney, 1998;Blockley et al., 2005;Abbott et al., 2011,2013;Griggs et al., 2014). The tephra shard concentrations were initially counted at a low-resolution (i.e. non- contiguously for every 2e4 cm) and recounted at a higher resolution (i.e. contiguously for every 1 cm) around identiﬁed shard concentration peaks.

2.2.1. Tephra shard concentrations

The selected material underwent several separation steps in order to isolate tephra from other materials and further separate grains based on grain-size and density fractions of interest.

For the extraction of glass shards, sub-samples of ca. 0.5 g were freeze-dried and homogenized. Further, to remove any carbonate material, they were immersed in 6 ml of dilute (10%) hydrochloric acid (HCl) for ca. 24 h. Then, they were sieved using 125m^{m and} 80mm stainless steel test sieves and a 25mm nylon mesh sieve to separate the glass shards in three recommended size fractions Table 1

Summary of key tephra horizons previously identified in the Greenland ice core tephra framework (Bourne et al., 2013,2015) that are targeted in this study. For each horizon, the following information is provided: label (including depth), climatic event, geochemistry afterLe Bas et al. (1986)(TB¼Tholeiitic Basalt, TAB¼Transitional Alkali Basalt, Da¼Dacite, R¼Rhyolite) and most likely volcanic source. In addition, the identified tephra horizons of MD99-2284 are added for each stratigraphicallyfitting Greenland ice core tephra layer.

Greenland ice core tephra layer Climatic event

Geochemistry Volcanic source

Reference StratigraphicallyﬁttingMD99- 2284 Horizons

NGRIP NEEM GRIP

1950.50 m 1689.25 m e GI-5 TB Kverkfj€oll Bourne et al. (2015) A, B, C

1952.15 m 1690.35 m e GS-6 TAB Katla Bourne et al. (2015) A, B, C

1954.70 m e e GS-6 TAB Katla Bourne et al. (2015) A, B, C

1973.16 m 1702.40 m e GI-6 TB Kverkfj€oll Bourne et al. (2015) A, B, C

2009.15 m e e GI-7 Da/R Hekla Bourne et al. (2015) D

2064.35 m 1755.60 m 2195.45 m GI-8 TB Grímsv€otn Bourne et al. (2013, 2015) E, F

2065.65 m e e GI-8 TB Grímsv€otn Bourne et al. (2013, 2015) E, F

2065.80 m e e GI-8 TB Grímsv€otn Bourne et al. (2013, 2015) E, F

2066.95 m 1757.10 m 2197.45 m GI-8 TB Grímsv€otn Bourne et al. (2013, 2015) E, F

2071.50 m 1759.85 m 2201.50 m GS-9 TB Grímsv€otn Bourne et al. (2013, 2015) E, F

2073.15 m e e GS-9 TB Grímsv€otn Bourne et al. (2013, 2015) E, F

2078.01 m 1764.25 m 2207.00 m GS-9 TB Grímsv€otn Bourne et al. (2013, 2015) E, F

2085.80 m e e GS-9 TAB Katla Bourne et al. (2013, 2015) E, F

2100.65 m e e GS-10 TB Grímsv€otn Bourne et al. (2013, 2015) F

2101.55 m e e GS-10 TB Grímsv€otn Bourne et al. (2013, 2015) F

2103.98 m 1780.20 m 2227.15 m GS-10 TB Grímsv€otn Bourne et al. (2013, 2015) F

(5)

Fig. 2.Tephra shard concentration proﬁles and climatic parameters of MD99-2284 within the studied time-interval (ca. 32e40 ka b2k). The light grey highlighted areas represent the Greenland Interstadials (GI) 5e9. In thisﬁgure, all MD99-2284 results are plotted versus age (ka b2k) on the GICC05 time-scale based on the age model presented inSadatzki et al. (2019). Dark grey crosses and dotted lines indicate 10 out of 11 ARM/SST-based tuning-points used for the latter age model. a) Anhysteretic remanent magnetization (ARM) (dark grey) and reconstructed near-surface temperature based on planktic foraminiferal assemblages (SST) (red) (Dokken et al., 2013). b) NGRIP ice cored¹⁸O plotted as 25 point

(6)

(Abbott et al., 2018b).

The separation of the smallest size fraction (i.e. 25-80 m^{m) is} important as it enhances the identiﬁcation ofﬁne-grained deposits (Davies, 2015). This fraction was separated into different density fractions using a heavy liquid separation technique originally developed for terrestrial sediments (Hodder and Wilson, 1976;

Turney, 1998; Blockley et al., 2005). Sodium polytungstate was prepared to create heavy liquids of both 2.3 gcm³and 2.5 gcm³, which was then used to isolate the rhyolitic and basaltic glass shards in density fractions of 2.3 gcm³ to 2.5 gcm³and >2.5 gcm³, respectively. All>2.5 gcm³fractions potentially containing basaltic glass shards underwent magnetic separation to aid in their isolation and identiﬁcation as it was previously demonstrated to be a valuable additional step within marine cryptotephra studies in the North Atlantic (Griggs et al., 2014;Abbott et al., 2018b). From the oven-dried >2.5 gcm³ fraction, the paramagnetic basaltic material was separated from the non-magnetic minerogenic material using a Frantz Magnetic separator and the settings ofGriggs et al. (2014)(i.e. tilt¼ 15, slope¼22.5, current¼0.85 nA).

Finally, all the different size fractions from each sample were mounted separately in Canada Balsam on microscope slides and inspected using optical light microscopy for their tephra shard concentrations (#/g dry sediment) by identifying basaltic (brown- colored) and rhyolitic (transparent) glass shards.

2.2.2. Ice rafted debris concentrations

To obtain a better understanding of secondary transport mechanisms, an IRD record was also constructed. Such a record has proven to be important when ice-rafting processes might compromise the integrity of tephra layers in high-latitude settings during glacial periods (Brendryen et al., 2010;Griggs et al., 2014;

Abbott et al., 2018b). In particular, tephra shards deposited onto ice- sheets can be transported to marine core sites after calving and rafting of icebergs. The ice-rafting might impart a temporal delay between eruption and ﬁnal deposition of several millennia and thereby, hamper the isochronous nature of tephra deposits (Brendryen et al., 2010). Depending on the ice-sheet calving time, the ice-rafting might deposit shards larger than those typically associated with air fall as well as a compilation of tephra originating from different volcanic eruptions. The latter results in a heterogeneous geochemical composition (Ruddiman and Glover, 1972;

Lackschewitz and Wallrabe-Adams, 1997;Brendryen et al., 2010;

Abbott et al., 2011). Hence, in addition to the geochemical composition of a tephra horizon, an important criterion to recog- nize ice-rafted tephra horizons is a co-varying IRD record (Lackschewitz and Wallrabe-Adams, 1997;Davies et al., 2014). The IRD concentrations (#/g dry sediment) were calculated by counting the IRD fragments (>150 mm), excluding any volcanic material, continuously for every 2 cm.

2.2.3. Geochemical analysis of cryptotephra deposits

Samples showing a maximum peak in tephra shard concentrations alongside no increased IRD input for the same sample depth were chosen for geochemical analysis. Prior to embedding the sample material in epoxy resin on frosted microprobe slides (2448 mm) the same separation steps as previously described (see 2.2.1.) were applied. Subsequently, the epoxy resin was ground using silicon carbide paper to expose the surface of the glass shards

and polished with diamond polycrystalline suspension (3, 1 and 0.5mm) and alumina powder to provide an unscratched surface for geochemical analysis (Griggs et al., 2014;Abbott et al., 2018b).

The geochemical characterization of individual glass shards was conducted at the Tephrochronological Analytical Unit at the Uni- versity of Edinburgh using electron-probe microanalysis (EPMA). A Cameca SX-100 electron microprobe withﬁve vertical wavelength dispersive spectrometers was operated over several analytical periods to measure oxide concentrations (wt. %) of ten major elements. All measurements were performed using the operating conditions outlined inHayward (2012). Depending on the grain- size of the samples, either a 3 or 5mm beam diameter was used.

To determine the instrumental drift within analytical sessions as well as assess the precision and accuracy of analysed samples, the secondary standards of Lipari Obsidian and BCR2g basalt were analysed at the beginning and end of each day. From each sample, ca. 20 individual shards were analysed to obtain a geochemical characterization that allows for a thorough assessment of the isochronous nature of the respective tephra layer. All analyses with total oxide values<97% were rejected. In order to compare the new marine tephra data of this study with the previously published data from the Greenland ice core tephra framework (Bourne et al., 2013, 2015), the major element data were normalized to an anhydrous basis (i.e. 100% total oxides). The latter was done due to the different levels of post-depositional hydration that might occur between the different depositional environments (Abbott et al., 2011;Pearce et al., 2014). In particular, tephra shards preserved in the marine realm are known to be susceptible toﬂuctuating levels of post-depositional hydration (Wallrabe-Adams and Lackschewitz, 2003;Abbott et al., 2011). The full geochemical results and secondary standard data are provided in the Supplementary Data.

In order to determine potential matches between marine tephra deposits and Greenland ice core tephra horizons, the geochemical populations of stratigraphicallyfitting horizons were both visually and statistically compared. The similarity coefficient (SC) was used to compare major elements from different datasets by calculating their similarity (Borchardt et al., 1972;Beget et al., 1992;Hunt et al., 1995). As the lower precision of low concentration elements might influence the coefficient, elements with concentration below 1 wt.

% were not considered (Hunt et al., 1995). Consequently, only seven major elements were taken into consideration for the SC calculation. According toBeget et al. (1992), values between 0.95 and 1 indicate that both datasets are identical. However, here we have followed the recommendation ofAbbott et al. (2018a)who applied a stricter approach as for some of the Icelandic centers values higher than 0.95 might just indicate a common source and values higher than 0.97 are more indicative of an identical geochemical composition. Another statistical parameter used was statistical distance (D²) developed byPerkins et al. (1995,1998). This method uses the Euclidian distance between samples to determine the difference between the geochemical datasets. The critical value for testing D²values using 10 major elements, at the 99% conﬁdence interval, is 23.21 (10 degrees of freedom). D²values higher than the critical value indicate that the datasets being compared can be considered different and thus, the hypothesis that the samples are identical can be rejected (Pearce et al., 2008).

moving average (purple) (Andersen et al., 2004,2006;Seierstad et al., 2014) and volcanic horizons (vertical bars) and their volcanic systems identiﬁed byBourne et al. (2013,2015).

c) Ice rafted debris (IRD) concentrations counted in the>150mm size fraction (brown). The light brown shaded areas represent intervals with increased input of IRD. d-f) Basaltic (green) and rhyolitic (blue) tephra shard concentrations counted in 25e80mm, 80e125mm and>125mm size fractions for the selected depth-intervals: (1) 2108e2148 cm, (2) 2270e2350 cm, (3) 2645e2700 cm and (4) 3020e3180 cm. These depth-intervals are indicated in numbered green and blue boxes for the quantiﬁcation of basaltic and rhyolitic concentrations, respectively. Each peak concentration is highlighted by a soft red line and labeled as Horizon A-F (marine core depth). (For interpretation of the references to color in thisﬁgure legend, the reader is referred to the Web version of this article.)

(7)

3. Tephrostratigraphy

The new tephrostratigraphy is presented here through de- scriptions of the high-resolution tephra shard and IRD concentration proﬁles including the different glass shard size fractions versus the chronology ofSadatzki et al. (2019)(Fig. 2; see 3.1.). The iden- tiﬁed tephra horizons are assessed for their potential isochronous nature following the assessment protocol outlined byAbbott et al.

(2018b). Subsequently, to determine if the selected horizons can also act as independent time-parallel markers, the geochemical signatures are presented and compared with existing geochemical data from stratigraphically similar Greenland ice core tephra horizons (Bourne et al., 2013,2015) (see 3.2.).

3.1. High-resolution shard concentrations

In general, there is a pronounced difference between the tephra shard proﬁles of GI-5 to GI-7 and the tephra concentrations of GI-8/

GS-9 (Fig. 2def). Within the coarser size fractions, the shard concentrations of GI-5/GS-6 and GI-7/GS-8 hardly record any tephra (Fig. 2eef) despite a clear observation of, respectively, two and one concentration peak(s) in the 25e80mm fraction (Fig. 2d). For GI-6/

GS-7, one distinct tephra peak is recorded in all size fractions around the same depth, albeit slightly offset between the different size fractions (Fig. 2def). During GI-8/GS-9, however, the shard proﬁles of all three size fractions show a profoundly different pattern. In particular, compared to GI-8, the coarser-grained size fractions (i.e.>125m^{m and 80}e125mm) show throughout GS-9 much higher concentrations including multiple concentration peaks (Fig. 2eef). Yet, theﬁne-grained fraction (i.e. 25-80mm) records one pronounced concentration peak during GI-8 and the most distinctive of GS-9 towards the end of the selected depth- interval (Fig. 2d).

The IRD record generally displays high concentrations during the interstadials and low concentrations during the stadials (Fig. 2c). However, an extremely high IRD input is recorded during GS-9. These periods of increased IRD are indicative that icebergs reached the core site and potentially deposited tephra shards that are older in age compared to their current stratigraphic position in the sediment core (Brendryen et al., 2010; Griggs et al., 2014;

Abbott et al., 2018b). Any tephra shard concentration peak that co- occurs with an increased level of IRD is not further discussed (see 2.2.2.). This includes the high tephra shard concentrations in the coarser-grained size fractions observed during GS-9 (Fig. 2eef). In addition, depth-interval 4 stratigraphically correlates to several closely spaced tephra deposits recorded in the Greenland ice cores (Bourne et al., 2013,2015) (Fig. 2b). Within this depth-interval, the smallest size fraction (i.e. 25-80mm) is characterized by high shard concentrations including two distinct and multiple smaller peaks (Fig. 2d). Due to the complexity of this ash zone (see 2.1.1.) and the observed period of coeval IRD during GS-9, only the two distinct horizons are further investigated. The youngest tephra peak in this interval (i.e. at 3038e3039 cm) (Fig. 2d) falls extremely close to the previously identiﬁed tephra horizon presented bySadatzki et al.

(2019) (i.e. at 3040e3041 cm) (see 2.1.2.) and represents in all likelihood the same horizon. However, as there is a striking difference in shard concentration (i.e.>22 000 shards/g dry sediment for 3038e3039 cm versus >12 000 shards/g dry sediment for 3040e3041 cm) this more distinct tephra peak, only recorded in this high-resolution investigation, is further investigated in order to reﬁne this horizon’s stratigraphic position.

This study focuses on the following six layers, labeled as follows:

Horizon A (2109e2110 cm), Horizon B (2134e2135 cm), Horizon C (2322e2323 cm), Horizon D (2648e2649 cm), Horizon E (3038- 3039 cm) and Horizon F (3173e3174 cm) (Fig. 2). For each of these

horizons a comprehensive multiple grain geochemical analysis is executed and compared to the geochemical data of the Greenland ice core tephra deposits within broad stratigraphic windows.

3.2. Identiﬁed tephra horizons 3.2.1. Horizon A (2109e2110 cm)

Horizon A, stratigraphically deposited during GI-5 at 2109e2110 cm core depth, represents a rapid increase in shard concentration with a peak in the 25e80mm fraction (Figs. 2d and 3c). To better visualize the shard profiles, the different concentrations of this layer are plotted versus depth (Fig. 3aec). In particular, the profile of thefine-grained shards shows a sharp transition/flat bottom profile and, compared to the background shard concentrations of this particular depth-interval, a distinct tephra peak with a high concentration of>3100 shards/g dry sediment (Fig. 3c;

Table 2). However, this peak is not mirrored in the coarser-grained shard proﬁles with only 2 and 6 shards/g dry sediment recorded in the>125m^{m and 80}e125mm size fractions, respectively (Fig. 3aeb;

Table 2).

The geochemical composition of Horizon A is homogeneous tholeiitic basalt (Fig. 3e). The major element composition shows a geochemical signature characterized by ranges of 50.08e51.08 wt.

% SiO2, 4.54e5.39 wt. % MgO, 0.41e0.61 wt. % K2O and 3.10e3.53 wt.

% TiO₂ (Fig. 3eei). Overall, this compositional signature has a geochemical afﬁnity with Kverkfj€oll sourced material (Fig. 3f and i).

Three different deposits from the Greenland ice core tephra framework stratigraphicallyﬁt with Horizon A (Fig. 2b;Table 1). Of these, NGRIP 1950.50 m/NEEM 1689.25 m (GI-5) exhibits a tight and homogeneous Kverkfj€oll composition, whereas the other two deposits consist of Katla type material (Fig. 3g). Prior to GS-6, another Kverkfj€oll sourced deposit (i.e. NGRIP 1973.16 m/NEEM 1702.45 m (GI-6)) was identiﬁed byBourne et al. (2015)(Figs. 2b and 3g;Table 1). A comparative analysis of Horizon A and those four ice core deposits, both visually (Fig. 3eei) and statistically (Table 3), indicates that Horizon A is geochemically most similar to NGRIP 1950.50 m/NEEM 1689.25 m (GI-5) (SC¼0.98; D²¼0.93).

However, as NGRIP 1973.16 m/NEEM 1702.45 m (GI-6) also consists of Kverkfj€oll sourced material, Horizon A does also show some similarities to it (SC ¼0.97; D² ¼8.40) (Fig. 3g;Table 3). None- theless, the Al2O3 and TiO2 concentrations clearly allow for the discrimination of NGRIP 1950.50 m/NEEM 1689.25 m (GI-5) from NGRIP 1973.16 m/NEEM 1702.45 m (GI-6) (Fig. 3i).

In summary, the shard profile of Horizon A is defined by a distinct and relatively high tephra concentration peak not co- occurring with increased IRD and consisting of a homogeneous geochemistry. These combined characteristics are defined as a deposit type-2A (Abbott et al., 2018b) arguing for Horizon A to be most likely deposited isochronously, via primary air fall. In addition, the clear visual, statistical and stratigraphical match provides independent evidence for a correlation between Horizon A and NGRIP 1950.50 m/NEEM 1689.25 m (GI-5). The identification of this isochron (i.e. MD99-2284_2109.5 cm) allows that this marine sediment core depth is used as a time-parallel marker for the following synchronization exercise (see 4.). The correlation to the Greenland tephra allows an age of 32.463 ka b2k from the GICC05 time-scale to be ascribed to this marine sediment core depth (Table 3).

3.2.2. Horizon B (2134e2135 cm)

Within theﬁrst depth-interval (2108e2148 cm), a second shard concentration peak in the 25e80 mm fraction (Horizon B) is observed at 2134e2135 cm core depth during, stratigraphically, late GS-6 (Figs. 2d and 4c). The shard proﬁle of this layer is, compared to

(8)

Fig. 3.Summary of Horizon A (2109e2110 cm). a-c) Tephrostratigraphy for selected depth-interval 1 (2108e2148 cm) with basaltic tephra shard concentrations for the different size fractions. Note: high-resolution areas are visible when there is no space left between the green bars. d) IRD concentrations. e-i) Comparison between the major elements of Horizon A (green diamonds) and stratigraphicallyfitting Greenland ice core horizons from GI-5, GS-6 and GI-6 (open circles) (Bourne et al., 2015). All data are normalized to an anhydrous basis (i.e. 100% total oxides). Error bars represent 2 standard deviations of replicate analyses of the BCR2g reference glass. e) Total alkalis vs. silica diagram (Le Bas et al., 1986). The dashed line indicates the boundary between the alkaline and subalkaline/tholeiitic series (MacDonald and Katsura, 1964). f) SiO2vs. FeO/TiO2biplot. g) FeO/MgO vs. TiO2biplot. h) K2O vs. MgO biplot. i) Al2O3vs. TiO2biplot. The different geochemicalfields for Icelandic source volcanoes are based on glass analyses shown inAbbott et al. (2018a)and references within (here presented in plots f and i) and shown inBourne et al. (2015)(here presented in plot g). (For interpretation of the references to color in thisfigure legend, the reader is referred to the Web version of this article.)

Table 2

Summary of tephra horizons detected in MD99-2284 between ca. 32 and 40 ka b2k. For each horizon, the following information is provided: label (including depth), shard concentration per counted size fraction, climatic event based on the age model presented inSadatzki et al. (2019), deposit type based on the classiﬁcation scheme presented in Abbott et al. (2018b), geochemistry afterLe Bas et al. (1986)(TB¼Tholeiitic Basalt, TAB¼Transitional Alkali Basalt, R¼Rhyolite) and most likely volcanic source.

MD99-2284 Tephra horizon

Tephra shard concentration (#/g dry sed.) Climatic event Deposit type Geochemistry Volcanic source

>125mm 80e125mm 25e80mm

Horizon A (2109e2110 cm) 2 6 3 141 GI-5 type-2A TB Kverkfj€oll

Horizon B (2134e2135 cm) 6 62 2 414 GS-6 type-2B/2A* TAB Katla/Kverkfj€oll

Horizon C (2322e2323 cm) 10 331 10 535 GI-6 type-2A TB Kverkfj€oll

Horizon D (2648e2649 cm) 6 18 3 037 GI-7 type-2A R Katla

Horizon E (3038-3039 cm) 14 311 22 123 GI-8 type-2A TB Grímsv€otn

Horizon F (3173e3174 cm) 45 730 6 911 GS-9 type-2B TAB Mixed

(9)

Table 3

Summary of the statistical comparisons between the detected tephra horizons in MD99-2284 and those from stratigraphically corresponding tephra horizons within the Greenland ice core tephra framework (Bourne et al., 2013,2015). For each tephra horizon, the following information is provided: label (including depth and number of analysed shards (n)), stratigraphically corresponding ice core tephra horizons, their climatic event and age (Bourne et al., 2013,2015) as well as the statistical parameters (SC¼similarity coefficient (Borchardt et al., 1972;Hunt et al., 1995) and D²¼statistical distance (Perkins et al., 1995,1998) indicative of a geochemicalfit. For statistical distance comparison, a critical value of 23.21 (10 degrees of freedom, 99% confidence interval) was used. The grey highlighted rows indicate the isochrons established in this study that can be used as time-parallel markers for synchronization (see 4.). Note: ^data is not included in the calculation of the statistical parameters.

(10)

Horizon A, less characteristic of a single primary tephra-fall event (Figs. 3c and 4c). Therefore, in addition to the concentration peak, the geochemistry of three additional depths within this shard proﬁle (i.e. one depth located 5 cm above the peak and two depths located 5 and 10 cm below the peak) were analysed and visualized (Fig. 4eek). Within this study, the peak depth (i.e. 2134-2135 cm) is labeled as Horizon B whereas the additional geochemical data within this shard proﬁle are referred to by their respective core depths.

Horizon B has, within the fine-grained shard profile, a peak concentration of>2400 shards/g dry sediment with a tail in both directions (i.e. a gradational upward span of 5 cm and 10 cm downwards) (Fig. 4c and e;Table 2). The>125m^{m and 80}e125m^m size fractions show concentrations of 6 and 62 shards/g dry sediment, respectively, but, compared to thefine-grained shard profile, do not record the gradational up- and downward spread (Fig. 4aeb;

Table 2). Importantly, the IRD record shows, despite an exception 4 cm above the tephra concentration peak, generally low values for this interval and thus, no co-occurring IRD increase with Horizon B (Fig. 4d).

The major element analyses from Horizon B exhibit a bimodal composition predominantly consisting of alkaline basalt and to a lesser extent tholeiitic basalt (Fig. 4g; dark green labeled as MD99- 2284 2134e2135 cm). These two geochemical populations show a clear difference in their oxide concentrations (Fig. 4gek; dark green diamonds versus triangles). The dominant population of Horizon B (i.e. population 1) records SiO2concentrations of 48.07e52.22 wt.

%, MgO concentrations of 4.34e5.36 wt. %, K₂O concentrations of 0.87e1.23 wt. % and TiO2 concentrations of 3.83e4.33 wt. % (Fig. 4gek: dark green diamonds), whereas the minor population

(i.e. population 2) is characterized by 49.36e51.18 wt. % SiO2, 4.69e4.99 wt. % MgO, 0.51e0.59 wt. % K₂O and 3.28e3.92 wt. % TiO₂ (Fig. 4gek: dark green triangles). The shards grouped as populations 1 and 2 display afﬁnities to Katla and Kverkfj€oll sourced material, respectively (Fig. 4h and k: dark green diamonds and triangles).

The close stratigraphic relationship between Horizons A and B means that the same Greenland ice core horizons to which Horizon A was compared fall within the stratigraphical proximity of Horizon B (Fig. 2b;Table 1). These ice core layers show either a Kverkfj€oll or a Katla sourced geochemical signature (Fig. 4i; Table 1). A comparative analysis between population 1 of Horizon B and those four ice core deposits, both visually (Fig. 4gek; dark green diamonds) and statistically (Table 3), shows clearly that population 1 is geochemically most similar to NGRIP 1952.15 m/NEEM 1690.35 m (GS-6) (SC¼0.97; D²¼1.69). However, in the SiO2concentrations, a slight offset is observed between population 1 and NGRIP 1952.15 m/NEEM 1690.35 m (GS-6) (Fig. 4h). As the secondary standard analyses from the two analytical periods also show a consistent slight offset in SiO₂ (See Supplementary Data for all standard measurements), it is most likely that the offset for the samples are, at least partially, explained by this variation reported in the standard analyses. And thus, the latter is partly regarded as an analytical offset rather than a real one that would prevent the correlation of the two horizons. Furthermore, even though NGRIP 1954.70 m (GS-6) has a Katla sourced geochemical signature and stratigraphic similarities, there is no correlative link between this ice core layer and any of the single glass shard measurements from Horizon B (Fig. 4gek;Table 3). The latter ice core layer exhibits a

(11)

Fig. 4.Summary of Horizon B (2134e2135 cm). a-c) Tephrostratigraphy for selected depth-interval 1 (2108e2148 cm) with basaltic tephra shard concentrations for the different size fractions. Note: high-resolution areas are visible when there is no space left between the green bars. d) IRD concentrations. e) Cryptotephra shard proﬁle including the relative distribution of population 1 (black)&2 (grey) within peak and tails. f) Percentage of shards from population 1 (black) and 2 (grey) for the analysed samples versus depth. g-k) Comparison between the major elements of Horizon B (green) and stratigraphicallyﬁtting Greenland ice core horizons from GI-5, GS-6 and GI-6 (open circles) (Bourne et al., 2015).

From each depth of the marine samples, the major and minor populations (i.e. population 1 (diamonds) versus population 2 (triangles)) are plotted separately. All data are normalized to an anhydrous basis (i.e. 100% total oxides). Error bars are representing 2 standard deviations of replicate analyses of the BCR2g reference glass. g) Total alkalis vs. silica diagram (Le Bas et al., 1986). The dashed line indicates the boundary between the alkaline and subalkaline/tholeiitic series (MacDonald and Katsura, 1964). h) SiO2vs. FeO/TiO2

biplot. i) FeO/MgO vs. TiO2biplot. j) K2O vs. MgO biplot. k) Al2O3vs. TiO2biplot. The different geochemicalﬁelds for Icelandic source volcanoes are based on glass analyses shown in Abbott et al. (2018a)and references within (here presented in plots h and k) and shown inBourne et al. (2015)(here presented in plot i). (For interpretation of the references to color in thisﬁgure legend, the reader is referred to the Web version of this article.)

(12)

tighter geochemical population with slightly higher TiO2and MgO, and somewhat lower SiO₂ concentrations compared to the full compositional range of Horizon B, which is also conﬁrmed by the statistical parameters (Table 3). While population 1 (Fig. 4gek: dark green diamonds) is stratigraphically, visually and statistically linked to NGRIP 1952.15 m/NEEM 1690.35 m (GS-6) and decoupled from NGRIP 1954.70 m (GS-6), population 2 of Horizon B (Fig. 4gek:

dark green triangles) shows some afﬁnities, both visually and statistically (Table 3), to the Kverkfj€oll sourced ice core horizons.

Statistically, population 2 shows the strongest correlation (SC¼0.99; D²¼2.94) (Table 3) to the stratigraphically older horizon (i.e. NGRIP 1973.16 m/NEEM 1702.45 m (GI-6)).

As the profile of thefine-grained shards is characterized by a tail in both an up- and downward direction and the bimodal composition reflects a double population, Horizon B has characteristics akin to deposit type-2B (Abbott et al., 2018b). This deposit type suggests that Horizon B is potentially hampered in its ability to be correlated to individual ice core horizons and thus, requires some additional investigations (Abbott et al., 2018a, 2018b). A more comprehensive analysis of the geochemistry from the entire shard profile was executed in order to investigate if secondary transport and depositional processes compromised the isochronous integrity of this tephra horizon. The geochemistry was measured for one sample located 5 cm above and two samples 5 and 10 cm below Horizon B (Fig. 4gek: 3 different lighter green colored labels compared to Horizon B). Akin to the geochemical signature of Horizon B, these additional samples also contain both Katla and Kverkj€oll sourced material (Fig. 4gek; Table 3). However, it is observed that Horizon B has proportionally more population 1 shards (Katla sourced) compared to the other samples (Fig. 4eef;

Table 3). These results argue for a certain degree of bioturbative reworking and/or secondary in wash upward as well as bioturbative reworking and/or vertical migration of shards downward of the peak concentration (Fig. 4e) (Davies, 2015;Griggs et al., 2015;

Abbott et al., 2018b). These post-depositional processes, mixing a single depositional event with existing background tephra, explains the decreasing trend of the Katla-sourced shards away from Hori- zon B. Indeed, compared to the tight homogeneous Katla-sourced population 1, population 2, which shows a strong correlation to the stratigraphically older NGRIP 1973.16 m/NEEM 1702.45 m (GI- 6) horizons (Fig. 4gek;Table 3), exhibits a much wider geochemical range within the full compositional range of Kverkfj€oll sourced material. This wider scatter of shards likely reﬂects some hetero- geneity within the second population due to post-depositional processes. Despite the initial signs of a non-isochronous deposit, features such as no coeval IRD (Fig. 4d), a link with the existing ice core horizon in a similar stratigraphic position (Figs. 2 and 4;

Table 3), and the relative dominance of population 1 within the shard profile (Fig. 4eef) provides strong evidence to suggest that population 1 of Horizon B reflects an air fall deposition from a single volcanic eruption. However, population 1 of Horizon B was deposited in addition to a background signal (i.e. population 2) that shows strong characteristics of a secondary deposition and/or reworked material from another eruption. The clear stratigraphic, visual and statistical match between population 1 of Horizon B and NGRIP 1952.15 m/NEEM 1690.35 m (GS-6) confirms its potential use as a time-parallel marker for future synchronization (i.e. MD99- 2284_2134.5 cm) (see 4.). This correlation allows us to ascribe a GICC05 age of 32.522 ka b2k to the marine sediment core depth (Table 3).

3.2.3. Horizon C (2322e2323 cm)

During GI-6, a clear peak in shard concentration (Horizon C) is observed at 2322e2323 cm core depth (Figs. 2d and 5a-c). Horizon

C shows a similar shard profile between the 25e80 mm and 80e125mm size fractions with concentrations of>10 500 and>300 shards/g dry sediment, respectively (Fig. 5bec; Table 2). Within both grain-sizes, the shard profile shows a high and distinct tephra peak with up- and downward tails of approximately 5 cm in both directions. However, the >125 mm size fraction has a slightly different profile with a concentration peak of 10 shards/g dry sediment 2 cm below the peaks in the smaller size fractions (Fig. 5a;

Table 2). Different settling velocities through the ocean and/or heavier material moving through soft sediment might explain such a depth-offset (Enache and Cumming, 2006).

Horizon C is characterized by a homogeneous tholeiitic basaltic composition with major elements between 49.75 and 51.28 wt. % SiO₂, 3.98e5.02 wt. % MgO, 0.52e0.63 wt. % K₂O and 3.1e3.8 wt. % TiO2with the exception of two outliers (Fig. 5eei). This geochemical signature shows a strong afﬁnity to Kverkfj€oll sourced volcanic material (Fig. 5f and i).

Only one tephra deposit preserved in the Greenland ice core framework correlates, stratigraphically, to Horizon C (Fig. 2b;

Table 1). This ice core deposit (i.e. NGRIP 1973.16 m/NEEM 1702.45 m (GI-6)) has a homogeneous Kverkfj€oll geochemical composition (Fig. 5g). However, the ice core deposit of GI-5 (i.e.

NGRIP, 1950.50 m/NEEM 1689.25 m) also has a similar Kverkfj€oll sourced geochemistry (Fig. 5g). Therefore, Horizon C was compared, both visually (Fig. 5eei) and statistically (Table 3), to both ice core deposits. These results suggest that Horizon C has a better match with the glass composition of NGRIP 1973.16 m/NEEM 1702.45 m (GI-6) (SC¼0.97; D²¼1.78) (Fig. 5g;Table 3).

The pronounced peaks in shard concentration are distinct across all grain-size fractions, with a slight depth-offset noted for the highest size fraction. Horizon C is characterized by a clear peak in shard concentration not associated with IRD input and a homogeneous geochemical population, which argues for a single depositional event that is most likely dominated by a primary air fall deposition. This allows Horizon C to be defined as a deposit type-2A (Abbott et al., 2018b). Combining this attribution with the clear statistical, visual and stratigraphicalfit, Horizon C can be considered isochronous and is correlated to NGRIP 1973.16 m/NEEM 1702.45 m (GI-6). Due to the pronounced tephra peak within the 25e80 mm size fraction and the mechanisms that could have influenced the stratigraphic position of the heavier shards, the isochron is labeled as MD99-2284_2322.5 cm and used as a time- parallel marker for future synchronization (see 4.). Due to the correlation with NGRIP 1973.16 m/NEEM 1702.45 m (GI-6), an age of 33.686 ka b2k on the GICC05 time-scale is ascribed to the marine sediment core depth (Table 3).

3.2.4. Horizon D (2648e2649 cm)

Horizon D is recorded at a core depth of 2648e2649 cm, stratigraphically deposited during GI-7 and deﬁned as a clearly distinct concentration peak in the ﬁne-grained size fraction (Figs. 2d and 6c). A high concentration of >3000 shards/g dry sediment is recorded within the 25e80mm size fraction (Fig. 6c;

Table 2). However, in the coarser-grained fractions this deposit is not well-pronounced with tephra concentrations of 6 and 18 shards/g dry sediment in the>125m^{m and 80}e125mm size fractions, respectively (Fig. 6aeb;Table 2).

The geochemical composition of glass shards from the deposit displays a tight homogeneous rhyolitic population with the exception of two outliers (Fig. 6e). The major elements are characterized by ranges of 71.22e73.84 wt. % SiO2, 3.38e4.02 wt. % FeO, 3.61e3.90 wt. % K₂O and 0.85e1.57 wt. % CaO (Fig. 6eei). Although theses values are typical for rhyolitic tephras from Iceland, this compositional signature shows the strongest afﬁnity to Katla

(13)

sourced volcanic material (Fig. 6f, g and 6i). Notably, the rather high FeO concentrations allow for a discrimination between the different geochemicalﬁelds and indicate how this geochemistry overlaps with published data of the Vedde Ash (Fig. 6h).

From a stratigraphic point of view, one ice core tephra deposit falls within the proximity of Horizon D (Fig. 2b;Table 1). However, this deposit (i.e. NGRIP 2009.15 m (GI-7)) exhibits a heterogeneous dacite to rhyolite composition (Fig. 6eei). It is clear from the bivariate plots (Fig. 6eei), and statistically conﬁrmed (Table 3), that Horizon D does not correlate to this ice core tephra deposit (SC¼0.85; D²¼15.25). In addition, the geochemistry of Horizon D does not correspond to any other tephra horizons with a similar

stratigraphic position in either ice core or marine records.

Based on the distinct concentration peak, the homogeneity of the geochemical composition and no coeval IRD, Horizon D is classiﬁed as a deposit type-2A (Abbott et al., 2018b) showing all evidence of a primary air fall deposit with a clear isochronous nature. However, as there is no correlation with the known Greenland ice core tephra framework, Horizon D cannot be used as a time- parallel marker in this work. Nevertheless, it does offer the potential to be used for future correlations if it can be traced in other climate archives.

Fig. 5.Summary of Horizon C (2322e2323 cm). a-c) Tephrostratigraphy for selected depth-interval 2 (2270e2350 cm) with basaltic tephra shard concentrations for the different size fractions. Note: high-resolution areas are visible when there is no space left between the green bars. d) IRD concentrations. e-i) Comparison between the major elements of Horizon C (green diamonds) and stratigraphicallyﬁtting Greenland ice core horizons from GI-5, GS-6 and GI-6 (open circles) (Bourne et al., 2015). All data are normalized to an anhydrous basis (i.e. 100% total oxides). Error bars are representing 2 standard deviations of replicate analyses of the BCR2g reference glass. e) Total alkalis vs. silica diagram (Le Bas et al., 1986). The dashed line indicates the boundary between the alkaline and subalkaline/tholeiitic series (MacDonald and Katsura, 1964). f) SiO2vs. FeO/TiO2biplot. g) FeO/MgO vs.

TiO2biplot. h) K2O vs. MgO biplot. i) Al2O3vs. TiO2biplot. The different geochemicalﬁelds for Icelandic source volcanoes are based on glass analyses shown inAbbott et al. (2018a) and references within (here presented in plots f and i) and shown inBourne et al. (2015)(here presented in plot g). (For interpretation of the references to color in thisﬁgure legend, the reader is referred to the Web version of this article.)

(14)

3.2.5. Horizon E (3038-3039 cm)

Within the ﬁnal depth-interval selected in this study, an extremely high and distinct tephra peak with concentrations of

>22 000 shards/g dry sediment is observed in the 25e80m^{m size} fraction at 3038e3039 cm core depth (Figs. 2d and 7c;Table 2). This tephra layer is labeled Horizon E and stratigraphically deposited during GI-8. Within the high-resolution proﬁle, the peak clearly stands out with no signs of a gradational tail. However, only 14 and>300 shards/g dry sediment are recorded within the>125m^m and 80e125mm size fractions, respectively (Fig. 7aeb; Table 2).

Compared to the glass shard concentrations down-core, these numbers do not stand out and thus, the coarser-grained shard

proﬁles do not capture this tephra layer within the marine sediment core (Fig. 7aeb).

Horizon E has a tholeiitic basaltic composition with the exception of one outlier (Fig. 7e). The rather tight ranges in its major elements, with values of 48.69e50.32 wt. % SiO2, 4.92e5.92 wt. % MgO, 0.39e0.56 wt. % K2O and 2.94e3.26 wt. % TiO2, reﬂect a strong homogeneity (Fig. 7eei)and a strong afﬁnity with the Grímsv€otn volcanic region (Fig. 7f and i).

GI-8 is well-known for the widely dispersed FMAZ III deposit in the marine realm (e.g.Rasmussen et al., 2003; Wastegård et al., 2006) and the series of eruptions recorded in the Greenland ice cores (Bourne et al., 2013, 2015). Stratigraphically, Horizon E Fig. 6.Summary of Horizon D (2648e2649 cm). a-c) Tephrostratigraphy for selected depth-interval 3 (2645e2700 cm) with rhyolitic tephra shard concentrations for the different size fractions. Note: high-resolution areas are visible when there is no space left between the blue bars. d) IRD concentrations. e-i) Comparison between the major elements of Horizon D (blue diamonds) and stratigraphicallyﬁtting Greenland ice core horizons from GI-7 (open circles) (Bourne et al., 2015). All data are normalized to an anhydrous basis (i.e.

100% total oxides). Error bars are representing 2 standard deviations of replicate analyses of the Lipari Obsidian reference glass. e) Total alkalis vs. silica diagram (Le Bas et al., 1986).

The dashed line indicates the boundary between the alkaline and subalkaline/tholeiitic series (MacDonald and Katsura, 1964). f) K2O vs. TiO2biplot. g) FeO vs. CaO biplot. h) SiO2vs.

K2O biplot. i) FeO vs. K2O biplot. The different geochemicalﬁelds for Icelandic source volcanoes are based on glass analyses shown inJennings et al. (2014)and references within (here presented in plots f, g and i). The geochemical envelopes for Borrobol and Vedde Ash are taken fromKoren et al. (2008)(here presented in plot g). (For interpretation of the references to color in thisﬁgure legend, the reader is referred to the Web version of this article.)

(15)

corresponds to several tephra deposits (Fig. 2b;Table 1). With the exception of NGRIP 2085.80 m (GS-9), consisting of Katla sourced volcanic material, all of these ice core deposits exhibit geochemical homogeneity and Grímsv€otn-like compositions (Fig. 7g;Table 1).

However, based on their TiO2content, the tephra deposits from GI- 8 (Fig. 7gei; yellow) can be discriminated from those deposited during GS-9 (Fig. 7gei; blue). A visual (Fig. 7eei) and statistical (Table 3) comparison of Horizon E and the GI-8/GS-9 ice core deposits shows that, despite the fact that the GS-9 ice core tephra layers are part of the wider Grímsv€otn-sourced geochemical en- velope, they do not correlate to Horizon E, neither visually (Fig. 7eei; blue) nor statistically (Table 3). Contrary, the comparison

highlights strong similarities between Horizon E and the, stratigraphically consistent, GI-8 ice core tephra layers (Fig. 7eei; yellow). Nonetheless, deﬁning a correlation between Horizon E and an individual GI-8 ice core horizon is very difﬁcult. In particular, when eliminating any offset data point (i.e. non-Grimsv€otn sourced data points inFig. 7eei) consistently for each ice core horizon, none of those four GI-8 horizons correlate separately to Horizon E with a SC>0.97 (Table 3). Therefore, it is not possible to disentangle the four separate GI-8 ice core horizons.

The distinct concentration peak within the 25e80m^{m fraction} shows strong evidence for a single depositional event. Additionally, the geochemical homogeneity and lack of IRD supports the Fig. 7.Summary of Horizon E (3038-3039 cm). a-c) Tephrostratigraphy for selected depth-interval 4 (3020e3180 cm) with basaltic tephra shard concentrations for the different size fractions. Note: high-resolution areas are visible when there is no space left between the green bars. d) IRD concentrations. e-i) Comparison between the major elements of Horizon E (green diamonds) and stratigraphicallyﬁtting Greenland ice core horizons (NGRIP) from GI-8 and GS-9 (open circles) (Bourne et al., 2013). All data are normalized to an anhydrous basis (i.e. 100% total oxides). Error bars are representing 2 standard deviations of replicate analyses of the BCR2g reference glass. e) Total alkalis vs. silica diagram (Le Bas et al., 1986).

The dashed line indicates the boundary between the alkaline and subalkaline/tholeiitic series (MacDonald and Katsura, 1964). f) SiO2vs. FeO/TiO2biplot. g) FeO/MgO vs. TiO2biplot.

h) K2O vs. MgO biplot. i) Al2O3vs. TiO2biplot. The different geochemicalﬁelds for Icelandic source volcanoes are based on glass analyses shown inAbbott et al. (2018a)and references within (here presented in plots f and i) and shown inBourne et al. (2015)(here presented in plot g). (For interpretation of the references to color in thisﬁgure legend, the reader is referred to the Web version of this article.)